Large-scale outflows in luminous QSOs revisited: The impact of beam smearing on AGN feedback efficiencies

25 

Full text

(1)

A&A 594, A44 (2016)

DOI: 10.1051/0004-6361/201527992

Astronomy

&

©c ESO 2016

Astrophysics

Large-scale outflows in luminous QSOs revisited

The impact of beam smearing on AGN feedback efficiencies

B. Husemann

1

, J. Scharwächter

2, 3

, V. N. Bennert

4

, V. Mainieri

1

, J.-H. Woo

5

, and D. Kakkad

1

1 European Southern Observatory, Karl-Schwarzschild-Str. 2, 85748 Garching b. München, Germany

e-mail: bhuseman@eso.org

2 LERMA, Observatoire de Paris, PSL, CNRS, Sorbonne Universités, UPMC, 75014 Paris, France 3 Gemini Observatory, Northern Operations Center, 670 North A’ohoku Place, Hilo, HI 96720, USA 4 Physics Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA 5 Department of Physics and Astronomy, Seoul National University, 151-742 Seoul, Korea

Received 17 December 2015 /Accepted 26 May 2016

ABSTRACT

Context. Feedback from active galactic nuclei (AGN) is thought to play an important role in quenching star formation in galaxies. However, the efficiency with which AGN dissipate their radiative energy into the ambient medium remains strongly debated. Aims. Enormous observational efforts have been made to constrain the energetics of AGN feedback by mapping the kinematics of the ionized gas on kpc scale. We study how the observed kinematics and inferred energetics are affected by beam smearing of a bright unresolved narrow-line region (NLR) due to seeing.

Methods. We re-analyse optical integral-field spectroscopy of a sample of twelve luminous unobscured quasi-stellar objects (QSOs) (0.4 <z<0.7) previously presented in the literature. The point-spread function (PSF) for the observations is directly obtained from the light distribution of the broad Hβline component. Therefore, we are able to compare the ionized gas kinematics and derived energetics of the total, truly spatially extended, and unresolved [O

iii

] emission.

Results. We find that the spatially resolved [O

iii

] line width on kpc scales is significantly narrower than the one before PSF deblend­ ing. The extended NLRs (ENLRs) appear intrinsically offset from the QSO position or more elongated which can be interpreted in favour of a conical outflow on large scales while a spherical geometry cannot be ruled out for the unresolved NLR. We find that the kinetic power at 5 kpc distance based on a spherical model is reduced by two orders of magnitude for a conical outflow and one order of magnitude for the unresolved NLR after PSF deblending. This reduced kinetic power corresponds to only 0.01−0.1 per cent of the bolometric AGN luminosity. This is smaller than the 5−10% feedback efficiency required by some cosmological simulations to reproduce the massive galaxy population. The injected momentum fluxes are close or below the simple radiation-pressure limit Lbol/c

for the conical outflow model for the NLR and ENLR when beam smearing is considered.

Conclusions. Integral-field spectroscopy is a powerful tool to investigate the energetics of AGN outflows, but the impact of beam smearing has to be taken into account in the high contrast regime of QSOs. For the majority of observations in the literature, this has not been addressed carefully so that the incidence and energetics of presumed kpc-scale AGN-driven outflows still remain an unsolved issue, from an observational perspective.

Key words. ISM: jets and outflows – galaxies: active – quasars: emission lines – techniques: imaging spectroscopy

1. Introduction

Feedback from active galactic nuclei (AGN) has become a key ingredient in numerical simulations and semi-analytic models of galaxy evolution to suppress star formation at the highest stellar masses, which appears necessary to recover the properties of the local galaxy population (e.g. Di Matteo et al. 2005; Bower et al. 2006; Croton et al. 2006; Somerville et al. 2008; Schaye et al. 2015). However, the mechanism(s) with which the released en­ ergy of AGN is dissipated to the surrounding interstellar medium of the host galaxy is poorly constrained by observations so far. One popular scenario for AGN feedback is a large-scale outflow where the AGN energy is sufficient to expel a large fraction of the gas from the host galaxy (e.g. Silk & Rees 1998). Thereby, the AGN is reducing the available gas reservoir and the star for­ mation activity becomes greatly suppressed.

The existence of high-velocity AGN-driven gas outflows has been confirmed by X-ray observations of ultra-fast outflowing

material in the circumnuclear region (e.g. Tombesi et al. 2010; Gofford et al. 2015). Also, broad-absorption line (BAL) AGN display outflowing gas with velocities of a few 1000 km s−1 in

(2)

blue wing that is interpreted as a genuine signature for an extended outflow (e.g. Heckman et al. 1981; Boroson 2005; Komossa et al. 2008; Mullaney et al. 2013). These high-velocity outflows are well resolved in very nearby Seyfert galaxies on

<1 kpc scales (e.g. Crenshaw & Kraemer 2000; Rice et al. 2006; Storchi-Bergmann et al. 2010; Fischer et al. 2013).

Currently, great efforts are being made to investigate the properties of large-scale high-velocity outflows in the most luminous AGN, that is quasi-stellar objects (QSOs). They are expected to show the strongest outflows if these are driven by the AGN radiation (e.g. Silk & Rees 1998; King 2003; Hopkins et al. 2010; Faucher-Giguère & Quataert 2012). With the advent of optical and near-IR long-slit or integral-field unit (IFU) spectrographs on 8 m class telescopes, the ENLR kinematics has been mapped on kpc scale for lu­ minous QSOs at low redshift z < 1 (e.g. Fu & Stockton 2009; Greene et al. 2011, 2012; Rupke & Veilleux 2011; Villar-Martín et al. 2011; Husemann et al. 2013b; Liu et al. 2013b, 2014, 2015; Harrison et al. 2014; McElroy et al. 2015; Humphrey et al. 2015; Villar-Martín et al. 2016; Karouzos et al. 2016) and high redshift z > 1 (e.g. Cano-Díaz et al. 2012; Harrison et al. 2012; Brusa et al. 2015; Cresci et al. 2015; Carniani et al. 2015). The majority of those studies report out­ flows on several kpc scales in almost all QSOs as indicated by broad and/or blue-shifted [O

iii

] emission lines.

Many of the spectroscopic QSO observations are focussed on obscured (type II) QSOs. Obscured QSOs lack the bright point-like power-law continuum of the accretion disc and the emission of the broad-line region (BLR) that are prominent in unobscured (type I) QSOs. This difference has been explained by the incli­ nation of a toroidal-like obscuring structure with respect to our line-of-sight in the unification model of AGN (e.g. Antonucci 1993). In this model the NLR is located outside the obscuring structure and can be seen in both types of QSOs. Given the high gas density and radiation field close to the AGN, the [O

iii

] emis­ sion lines from the NLR on scales of «1 kpc can outshine the ENLR on host galaxy scales by a factor of a few depending on the size of the ENLR and physical resolution of the observa­ tion. Therefore, for the purpose of our study we define the NLR and ENLR as the spatially unresolved and resolved emission, respectively.

It is therefore crucial to characterize the point-spread func­ tion (PSF) for QSO observations in order to separate the emis­ sion from the compact NLR and the contributions from the ENLR. The ability to achieve such a separation depends strongly on the spatial resolution. Characterizing the PSF is particularly challenging for spectroscopic observations of obscured QSOs, because the slit or IFU usually do not simultaneously cover a star. In these cases, an approximation of the PSF and its shape may be obtained from acquisition images (e.g. Hainline et al. 2013, 2014; Humphrey et al. 2015) or standard star observations (e.g. Liu et al. 2013a, 2014) taken close in time to the science observations. However, the actual PSF for the science observa­ tion can still be different due to time variability of the seeing and the tracking error of the telescope for significantly longer science exposures.

IFU spectroscopy of unobscured QSOs provides a way to reconstruct the PSF directly from the science data assuming that broad Balmer lines from the BLR are intrinsically unre­ solved (e.g. Jahnke et al. 2004). This technique was applied to various IFU observations of unobscured QSOs (Sánchez et al. 2004; Christensen et al. 2006; Husemann et al. 2008, 2013b, 2014; Carniani et al. 2015; Herenz et al. 2015; Liu et al. 2015) enabling the study of line diagnostics and kinematics across the

host galaxy without the apparent contamination of the bright un­ resolved NLR. Based on a large sample of luminous unobscured QSOs at z< 0.3 and luminous obscured QSOs, Husemann et al. (2013b), Villar-Martín et al. (2016), and Karouzos et al. (2016) reported a lack of high-velocity outflows on kpc scales after deblending the unresolved NLR and ENLR. This appears to be in direct contradiction to the result of various other groups for luminous obscured QSOs (e.g. Greene et al. 2011; Liu et al. 2013a, 2014; Harrison et al. 2014; McElroy et al. 2015). How­ ever, those studies did not separate kinematics of the NLR and ENLR given the difficulty of constraining the PSF. It is therefore unclear whether beam smearing, differences in the QSO feed­ back efficiency or even differences in the unobscured and ob­ scured QSOs sample selection are causing these discrepant conclusions.

In this paper, we systematically investigate the effect of the beam smearing on the measured kpc-scale kinematics of the [O

iii

] lines. In particular, Liu et al. (2013b) reported very large mass-outflow rates and kinetic power from IFU spectroscopy of the [O

iii

] line for a sample of luminous obscured QSOs at red-shift 0.4 <z< 0.7 . While the PSF for these observations cannot be reconstructed, the authors also presented a matched sample of unobscured QSOs in Liu et al. (2014) for which the BLR can be used as a PSF tracer. Here, we re-reduce and re-analyse the dataset of luminous unobscured QSOs from Liu et al. (2014) and compare the results with and without deblending the contribu­ tion from the unresolved NLR and ENLR. Thereby, we can ver­ ify how much the results on the ENLR geometry, the large-scaled ionized gas kinematics and associated AGN feedback efficiency are affected by beam smearing.

The paper is organized as follows. In Sect. 2 we describe the IFU data reduction and analysis including our QSO-host galaxy deblending scheme and emission-line measurements. This is followed by a comparison of various parameters on the ex­ tended ionized gas measurements before and after deblending the point-like and extended emission (Sect. 3). From the mea­ sured quantities we compute outflow energetics for two different outflow models in Sect. 4. We then discuss our results with previ­ ous observations and expectations for AGN feedback scenarios (Sect. 5). Finally, we close with a summary and our main conclu­ sions in Sect. 6. Throughout the paper we assume a concordance cosmological model with H0 = 70 km s−1 Mpc−1 , Ωm = 0.3, and

ΩΛ= 0.7.

2. Integral-field spectroscopy of luminous QSOs

2.1. The QSO sample and IFU data reduction

The QSO sample and optical integral-field observation we focus on in this paper are presented by Liu et al. (2014); here we briefly recap the main characteristics of the sample. The 12 QSOs were selected from the Shen et al. (2011) cat­ alogue to have (i) a minimum [O

iii

] luminosity of L[O iii] >

1042.7 erg s−1 to be comparable with the unobscured QSO selec­ tion in Liu et al. (2013a); (ii) a redshift range of 0.4 <z<0.7; (iii) a 1.4 GHz radio flux not exceeding f1.4 GHz < 10 mJy in the

NVSS or FIRST radio surveys to exclude radio-loud QSOs; and (iv) a high [O

iii

] equivalent width. In Table 1, we list some char­ acteristic parameters of the sample mainly taken from Liu et al. (2014), but we also compute additional parameters from the data itself, that is the broad Hβ line luminosity (LHβ) and the contin­

(3)

Table 1. Basic sample characteristics.

Identifier z log LOIII a

[erg s−1]

log LHβb

[erg s−1]

log L5100 c

[erg s−1]

log L8 µm d

[erg s−1]

f1.4 GHz

[mJy]

log P1.4 GHz

[W Hz−1]

Resolutione

SDSS J023342.57-074325.8 SDSS J030422.39+002231.8 SDSS J031154.51-070741.9 SDSS J041210.17-051109.1 SDSS J075352.98+315341.6 SDSS J080954.38+074355.1 SDSS J084702.55+294011.0 SDSS J090902.21+345926.5 SDSS J092423.42+064250.6 SDSS J093532.45+534836.5 SDSS J114417.78+104345.9 SDSS J221452.10+211505.1

0.4538 0.6385 0.6330 0.5492 0.4938 0.6527 0.5662 0.5749 0.5884 0.6864 0.6785 0.4752

43.1 42.8 42.9 43.5 42.6 43.2 42.7 43.1 43.0 43.2 43.3 42.8

42.6 43.9 42.6 43.7 42.7 43.8 43.1 43.2 43.5 43.1 43.5 43.1

44.8 45.7 45.0 45.4 44.8 45.6 44.9 45.2 45.4 45.1 45.2 44.9

45.0 45.9 45.8 45.8 44.6 45.7 45.0 45.6 45.5 45.3 45.2 45.1

<1.1

<0.8

<1.1 3.2

<1.0

<1.0

<1.0

<1.0

<1.1

<1.0

<1.0

<2.5

<23.8

<24.0

<24.1 24.5

<23.9

<24.1

<24.0

<24.0

<24.1

<24.2

<24.2

<24.2

0.56""/

3.2 kpc 0.55""/

3.8 kpc 0.56""/3.8 kpc

0.48""/

3.1 kpc 0.58""/

3.5 kpc 0.67""/

4.6 kpc 0.61""/

4.0 kpc 0.68""/

4.5 kpc 0.61""/

4.0 kpc 0.77""/

5.5 kpc 0.68""/

4.8 kpc 0.46""/

2.7 kpc

Notes. (a) Total [O

iii

] line luminosity from Liu et al. (2014). (b) Broad Hβline luminosity based on the QSO spectral modelling. (c) QSO continuum

luminosity at 5100 Å. (d) Continuum luminosity at 8 µm from Liu et al. (2014). (e) Angular and physical resolution of the GMOS IFU data measured

from the re-constructed broad HβPSF.

Observations of the QSO sample were taken with the Gemini multi-object spectrograph (GMOS: Allington-Smith et al. 2002) in IFU-mode at the Gemini-north telescope as part of pro­ gramme GN-2012B-Q-29 (PI: G. Liu). We retrieved the raw data and corresponding calibrations from the GEMINI science archive after the data became publicly available. The two-slit mode of the GMOS IFU provides a 5""× 7""target field-of-view

(FoV) that is contiguously sampled with 1000 hexagonal lenslets of 0"".2 in diameter. Additionally, 500 lenslets are packed into a

5""× 3.5"" FoV about 1"offset from the primary IFU field to si­ multaneously monitor the sky. The spectral range was chosen such that Hβ and [O

iii

] lines are simultaneously covered. Two different setups with the R400-G5305 grism (R ∼ 2000) in the i band are necessary to capture those important lines considering the redshift range in the sample and to avoid that lines falling in one of the gaps between the three charged-coupled devices (CCDs). Two 1620 s exposure were obtained for each QSO in the sample.

For the data reduction, we use the IFU data reduction package developed and extensively tested for the Calar Alto large integral field area (CALIFA) survey (Sánchez et al. 2012; Husemann et al. 2013a; García-Benito et al. 2015). Since CAL­ IFA uses the fibre-based IFU spectrograph PMAS (Roth et al. 2005), all reduction steps are almost identical except that the GMOS IFU samples the FoV contiguously and that the data is spread over three independent CCDs. The data reduction work-flow consists of standard tasks such as bias subtraction, cosmic-ray masking with PyCosmic (Husemann et al. 2012), fi­ bre tracing and fibre profile fitting using the continuum lamp ex­ posure, flexure correction, optimal fibre extraction, wavelength calibration using the attached arc lamp exposure, and relative wavelength-dependent fibre transmission correction using a con­ tinuum lamp. One important difference with respect to the data reduction of Liu et al. (2014) is that we re-sample the data into a datacube with 0"".2 rectangular spaxels. Over-sampling the native

data resolution with just two exposures does not provide addi­ tional information and would degrade the S/N per final spaxel. For the re-sampling, we assume that the hexagons effectively collect light within a circular aperture of 0.2""diameter and ap­ ply the “drizzle” resampling scheme (Fruchter & Hook 2002) to construct the final datacubes. During this re-sampling step, we simultaneously correct for the effect of atmospheric dispersion

by shifting the sample grid to account for the continuous shift in the relative position along wavelength.

Standard star observations are reduced in the same way as the science data to perform a relative spectrophotometric flux calibration along wavelength. Following Liu et al. (2014), we re­ trieve Sloan digital sky survey (SDSS, York et al. 2000) spec­ tra from DR10 (Ahn et al. 2014) and compare the GMOS spec­ tra with the SDSS ones to anchor our absolute flux calibration. Here, we simply extract spectra within a 3"" diameter centred

on the QSO and compare it directly with the SDSS DR10 spec­ tra since they were already re-scaled in flux to account for the aperture fibre losses. Then we determine the photometric scale factor compared to the SDSS spectra as the median of the ratio between the two spectra. We show the SDSS and the matched aperture GMOS spectra in Fig. 1. Although the spectrophoto­ metric calibration of SDSS spectra are considered very accurate, the GMOS and SDSS data are taken a few years apart so that in­ trinsic variability of AGN in the continuum and broad lines will lead to systematic uncertainties in our adopted absolute photo­ metric calibration.

2.2. Spatially-resolved [O III] emission-line analysis

(4)

Fig.1.Comparison of the GMOS 3""

aperture spectra (black line) with the corresponding SDSS spectrum (red line). The wavelength range is limited to the rest-frame wavelength range between 4500 and 5500 Å. An absolute photometric calibration of the GMOS data based on the median of the ratio of spectra over the common wavelength range. The spectra are normalized in flux density so that the peak in the [O

iii

] λ5007 line is set to one in the SDSS spectrum. A large part of the SDSS spectrum of SDSS J0304+0022 is masked as bad, as seen as the linear interpolated region. In general, the relative GMOS flux calibration across the wavelength range is consistent with the SDSS spectra at a <10% level.

Here, we specifically want to test how much the light from the unresolved NLR blends with the ENLR in IFU observations, altering the spatially resolved line profiles and biasing the de­ rived quantities. Unobscured QSOs are ideal for this purpose since the broad emission-lines from the unresolved BLR pro­ vide an intrinsic measurement for the PSF of a given observa­ tion (e.g., Jahnke et al. 2004). This allows us to accurately de-blend spatially unresolved, NLR or QSO, and resolved, ENLR or host galaxy emission, in an empirical way. We will refer to this process as a NLR-ENLR deblending or QSO-host galaxy deblending. To make a fair comparison, we characterize the [O

iii

] λλ4960,5007 doublet line profile spaxel by spaxel in a consistent way before and after applying NLR-ENLR deblend­ ing as described below.

2.2.1. Mapping the total [O III] line profile

To characterize the spatially resolved [O

iii

] emission-line pro­ file, we follow the algorithm of Liu et al. (2013b) which consists of three basic steps: 1) removal of Fe

ii

and broad Hβ emission; 2) multi-component modelling of the [O

iii

] doublet line; and

3) non-parametric line shape measurements based on the best-fit model. The first step is achieved by modelling the QSO spec­ trum with a set of Gaussian profiles to separate the various emis­ sion line components in the spectral range (see Fig. 2). Then, we create a best-fit model only for the broad Hβ and Fe

ii

line components as well as the local AGN power-law continuum. This template spectrum is subtracted from each spaxel after proper matching in flux. The residual is a pure narrow Hβ plus [O

iii

] emission-line datacube. According to Liu et al. (2013b), we model the [O

iii

] doublet lines as a superposition of up to three independent Gaussian systems coupled in their intrinsic flux ratio (1:3, Storey & Zeippen 2000), redshift and line dis­ persion. A Levenberg-Marquardt minimization algorithm with reasonable starting values is applied to determine the best-fit pa­ rameters per spaxel. Any model with an increased number of free parameters will provide a better fit. Based on a statistical F-test we decide whether the increased number of free parameters sig­ nificantly improved the χ2 in excess of what is expected from

(5)

7500

7600

7700

7800

7900

8000

8100

0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

flux density

7500

7600

7700

7800

7900

8000

8100

wavelength [ ]

0.04

0.02

0.00

0.02

0.04

residual

SDSSJ0924+0642

Fig. 2. Example of the broad Hβ and Fe

ii

emission-line subtrac­ tion without QSO-host deblending for SDSS J0924+0642. The ob­ served spectrum (black line) and our best-fit model (red line) are shown in the upper panel. The best-fit model consisting of the broad Hβand Fe

ii

λλ4948,5017 plus continuum is represented by the blue line with the corresponding residual spectrum of the narrow Hβ and [O

iii

] λλ4960,5007 indicated by the green line. The residuals of the to­ tal model are shown in thepanelbelow. Details of the assumed model are given in the main text.

width at the 80 per cent quantile of the line flux (W80), the line

asymmetry (A) and line kurtosis (K) following the formula pre­ sented in Liu et al. (2013b).

2.2.2. Mapping the ENLR [O III] line profile

We repeat the entire analysis process again, but now replacing step 1) with a spectral QSO-host galaxy deblending scheme to separate the apparently unresolved (NLR) and resolved (ENLR) [O

iii

] line emission. The explicit modelling and subtraction of the broad Hβ and Fe

ii

emission lines is not necessary, because those emission lines originate from the BLR and are intrinsically unresolved emission associated with the QSO spectrum and au­ tomatically subtracted during QSO-host galaxy deblending pro­ cess as shown in Fig. 3. For the QSO-host deblending we adopt an iterative algorithm implemented in the public software pack­ age

qdeblend

3D (Husemann et al. 2013b, 2014).

qdeblend

3D first re-constructs the PSF of the observations from the strength of the broad emission line, the Hβ line in this case. In the first iteration, the brightest spaxel which is dom­ inated by the QSO light is scaled according to the PSF and subtracted from each spaxel. The central spaxel contains not only spatially unresolved emission from the QSO, but also a fraction of host galaxy emission including the ENLR. The al­ gorithm iteratively removes the host galaxy contribution based on the average surface brightness of the residual host galaxy emission (Σhost) in the spaxels around the central QSO spaxel

(Σcore) after each iteration. A scale factor Σcore/Σhost is applied to

scale the brightness towards the centre, which clearly depends on the intrinsic surface brightness profile of the extended host galaxy/ENLR emission. It can reasonably vary only between a factor of one (constant surface brightness) and a factor corre­ sponding to purely unresolved emission depending on the PSF and GMOS sampling (see Fig. 4).

1.0

0.5

0.0

0.5

1.0

x

[arcsec]

1.0

0.5

0.0

0.5

1.0

y

[arcsec]

7500

7600

7700

7800

7900

8000

8100

wavelength [ ]

0.005

0.000

0.005

0.010

0.015

0.020

0.025

flux density [a.u.]

Fig.3.Upperpanel: example PSF estimated from the intensity of the broad Hβline for SDSS J0924+0642. The white circle indicates the FWHM of the seeing. We highlight two spaxels with a black and red square for which we show the spectra from the datacube in the lower panels. The red spaxel is 0"".8 (5.8 kpc) away from the QSO position.

Lower panel: spectra from the two spaxels highlighted in the upper panel. The central spaxel spectrum is scaled in to match in the inte­ grated broad Hβflux of outer spectrum. The difference between the spectra are indicated by the blue line and shows that both spectra are identical in shape except of a constant continuum offset across the wave­ length range. Any apparent emission line contribution in the red spaxel is simply due to beam smearing of an unresolved source even for the forbidden [O

iii

] line from the NLR.

(6)

1.0 0.5 0.0 0.5 1.0

arcsec

0.0 0.2 0.4 0.6 0.8 1.0

Intensity [a.u.]

host

core

host

0.5"

0.55"

0.6"

0.65"

0.7"

0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80

seeing [arcsec]

0.8 1.0 1.2 1.4 1.6 1.8 2.0

scale factor (

Σ

core

/

Σ

host

)

forbidden region

allowed region

Fig.4.Upperpanel: intensity profile for various seeing conditions as­ suming a Gaussian shape. The region of the central spaxel (core) and the adjacent host galaxy spaxel (host) are indicated by shaded areas for the GMOS sampling. Lower panel: corresponding scale factors for a point-source as a function seeing is shown by the black line. Higher scale factors would imply surface brightness distributions steeper than point-like sources (red shaded area).

all the QSOs from the total line profile and from the ENLR only after applying the QSO-host deblending.

3. Quantifying the impact of an unresolved NLR on ENLR measurements

3.1. Comparison with the original measurements

Since we performed a completely independent re-analysis, from the data reduction to the data analysis, compared to the work by Liu et al. (2014), we first want to test whether we recover their measurements if we follow their methods as closely as possible. In Fig. 6, we show a comparison of the [O

iii

] line width (W80),

maximum velocity range (Δv), the size of the ENLR (Rint) and

the power-law slope of the total [O

iii

] surface brightness pro­ file (I[O iii](R) ∼ R−η) over the range 1""−2.5"" from the QSO as

measured by Liu et al. (2014) and our own measurement from total [O

iii

] line maps before deblending.

We find that our measurements are in good agreement with the values reported by Liu et al. (2014). Systematic differences are less than 20% in all cases. Our measurements for W80 are

slightly smaller by 13 ± 10% and also the maximum velocity range Δv is smaller by 15 ± 17 per cent. For the latter, the rms is significant, which is caused by the systematic uncertainties on the measurements of the radial velocities given that we have

independently re-reduced the entire dataset. The most signifi­ cant scatter is found for the outer radial [O

iii

] surface bright­ ness profile η. The reason for this is that the radius over which the slope is measured is not clearly defined in Liu et al. (2014). The arbitrary fitting range of 1""2.5"" along the major axis of

the ENLR that we adopt here may simply not reflect the origi­ nal prescription to measure this parameter. However, our mean value of (η)= 3.5 ± 0.6 is consistent with the measurements for the unobscured and obscured QSOs by Liu et al. (2014).

We can almost exactly reproduce the isophotal radius of the ENLR, based on the surface brightness of concentric annuli, with a rather small deviation of 5 ± 5%. However, when the surface brightness of individual spaxels is concerned we can also de­ fine a ENLR size based on the largest projected distance from the QSO position to a single spaxel above the same thresh­ old surface brightness. These ENLR sizes can be significantly larger than the azimuthally averaged isophotal radii reported by Liu et al. (2014) and may explain the apparent flattening of the ENLR size – QSO luminosity relation at high QSO luminosities (Liu et al. 2014; Hainline et al. 2014). It is beyond the scope of this paper to investigate this in detail.

In the following, we study how much the measurements change after separating the compact unresolved NLR from the ENLR. After the tests discussed above, any difference we find can be unambiguously attributed to the effect of beam smearing. Given the high S/N of the data we do not report uncertainties on measured quantities in Table 2 which exhibits measurement er­ rors of less than 0.1 dex. Systematic uncertainties on the derived quantities will dominate the error budget by more than an or­ der of magnitude so that we can safely ignore the measurement errors.

3.2. Surface brightness distribution

We observe some subtle but important differences between the [O

iii

] surface distribution for some objects, after subtracting the compact unresolved NLR contribution. Here, we quantify the changes by means of a few important parameters. From the flux maps, we compute the total and ENLR [O

iii

] flux ( f[O iii])

within the GMOS FoV from which we define a contrast ratio as C= fENLR/( fNLR + fENLR); the fraction of the ENLR to the total

flux. In addition, we compute the flux-weighted centroid and el­ lipticity within a 2""×2""sub-frame centred on the QSO position. From the flux-weighted centroid, we infer the apparent distance to the QSO position (dQSO), defined as the flux-weighted centre

of the broad Hβ distribution. All those measurements are sum­ marized in Table 2.

We find that the contrast ratio C spans a large range across the sample. In four QSOs the ENLR contributes more than 50% to the total [O

iii

] emission, whereas the ENLR contributes less than 10% in the most extreme case. As expected, the changes in the ENLR surface brightness distribution appear marginal if C> 0.5. Remarkably, the distance between the peak in the sur­ face brightness distribution and the QSO position dQSO is gen­

(7)

2 1 0 1 2

y

[arcsec]

SDSS J0233-0743

3 2 1 0 1 2 3

2 1 0 1 2

y

[arcsec]

1e-01 1 10 100

Σ[OIII][10−16erg/s/cm2/arcsec2]

3 2 1 0 1 2 3

500 250 0 250 500

vmedian[km/s]

3 2 1 0 1 2 3

0 400 800 1200 1600 2000

W80[km/s]

3 2 1 0 1 2 3

0.5 0.0 0.5

A

3 2 1 0 1 2 3

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

K

2 1 0 1 2

y

[arcsec]

SDSS J0304+0022

2 1 0 1 2 3 4

2 1 0 1 2

y

[arcsec]

2 1 0 1 2 3 4 2 1 0 1 2 3 4 2 1 0 1 2 3 4 2 1 0 1 2 3 4

2 1 0 1 2

y

[arcsec]

SDSS J0311-0707

2 1 0 1 2 3 4

2 1 0 1 2

y

[arcsec]

2 1 0 1 2 3 4 2 1 0 1 2 3 4 2 1 0 1 2 3 4 2 1 0 1 2 3 4

2 1 0 1 2

y

[arcsec]

SDSS J0412-0511

2 1 0 1 2 3 4

∆x [arcsec]

2 1 0 1 2

y

[arcsec]

2 1 0 1 2 3 4

(8)

Table 2. Basic parameters inferred for the ENLR before and after the deblending process.

Name f[O iii] a

[10−16 erg s−1 cm−2]

Cb d

QSO c

[kpc]

ed R

max e

[kpc]

ηf W

80g

[km s−1]

NLR ENLR tot ENLR tot ENLR tot ENLR tot ENLR tot ENLR NLR

SDSS J0233-0743 154 122 0.44 0.199 0.455 0.18 0.43 17.1 16.1 2.1 1.4 471 357 568 SDSS J0304+0022 43 21 0.33 0.227 1.402 0.04 0.20 7.4 7.4 3.8 3.1 1307 441 1627 SDSS J0311-0707 53 23 0.31 0.331 1.458 0.08 0.16 10.0 9.7 4.0 2.8 1001 774 1072 SDSS J0412-0511 512 83 0.14 0.039 0.553 0.09 0.11 15.5 18.3 3.4 2.0 1204 885 1362 SDSS J0753+3153 43 19 0.30 0.090 0.360 0.05 0.02 7.1 8.6 3.9 3.5 259 230 265 SDSS J0809+0743 120 42 0.26 0.182 2.366 0.10 0.24 22.8 22.8 3.1 2.0 891 693 928 SDSS J0847+2940 19 47 0.72 0.750 1.304 0.03 0.05 12.0 11.0 2.8 2.4 362 320 910 SDSS J0909+3459 50 110 0.69 0.876 1.777 0.20 0.18 16.0 14.2 4.4 4.1 568 512 974 SDSS J0924+0642 85 11 0.11 0.100 0.883 0.06 0.04 9.5 5.9 3.5 −0.5 900 765 1064 SDSS J0935+5348 64 61 0.49 0.114 0.364 0.13 0.26 13.4 14.0 4.1 3.3 585 490 725 SDSS J1144+1043 63 114 0.65 0.384 0.807 0.23 0.33 20.2 20.2 3.6 3.3 676 574 900 SDSS J2214+2115 64 35 0.35 0.212 1.125 0.07 0.02 8.0 9.3 4.2 3.9 559 835 601

Notes. (a) Spatially integrated [O

iii

] line flux. (b) Contrast ratio defined as C=f

ENLR/( fNLR +fENLR). (c) Distance of the [O

iii

] flux-weighted centre

with respect to the broad Hβflux-weighted centre defining the QSO position. (d) Ellipticity of the [O

iii

] flux distribution. (e) Maximum porjected

size of the ENLR up to a local surface brightness of Σ[O iii] >10−15/(1 +z)4 erg s−1 cm−1 . ( f) Power-law slope of the radial [O

iii

] surface brightness

distribution between 1""

–2.5""

from the QSO. (g) Median [O

iii

] line width as described in the text over a radius of <0.6""

around the QSO.

0

400

800 1200 1600

W

80

[km

/

s] (this work)

0

400

800

1200

1600

W

80

[km

/

s]

(Liu)

0 100 200 300 400 500 600

v

max

[km

/

s] (this work)

0

100

200

300

400

500

600

v

max

[km

/

s]

(Liu)

0

1

2

3

4

5

η

(this work)

0

1

2

3

4

5

η

(Liu)

0

5

10

15

20

25

R

int

[kpc] (this work)

0

5

10

15

20

25

R

int

[kpc]

(Liu)

Fig. 6. Comparison of the total [O

iii

] line width (W80, upper left panel), maximum ve­

locity range (Δv, upper right panel), size of the ENLR (Rint, lower right panel), and the

power-law slope of the total [O

iii

] surface brightness profile (η, lower left panel) from the QSO as measured by Liu et al. (2014) and our re-analysis. We find good agreement be­ tween measurements with only a weak sys­ tematic offset and a scatter consistent with the intrinsic accuracy of measurements. The surface brightness profile slope η shows the greatest scatter because the actual range of the outer profile to measure η is not clearly specified in Liu et al. (2014), so that our mea­ surements are likely not exactly matching their methodology. For the ENLR we mea­ sure the isophotal radius at a surface bright­ ness (corrected for cosmological dimming) of

Σ[O iii] =10−15 erg s−1 cm−2 arcsec−2 (black cir­

cles) and the maximum projected distance to spaxels which exhibit the same threshold sur­ face brightness locally (grey symbols).

Something that is not strongly affected by the beam smear- with the studies of Hainline et al. (2013, 2014) who find that the ing is the size of the ENLR up to an intrinsic [O

iii

] surface ENLR maybe at most 0.1−0.2 dex smaller after a full PSF con­

−1 −1

brightness threshold of Σ[O iii] > 10−15/(1 +z)4 erg s cm , cor- volution of the [O

iii

] surface brightness distribution. Only for

rected for cosmological surface brightness dimming, which was SDSS J0924+0624 do we recover a substantially smaller ENLR defined in Liu et al. (2013a) and is an arbitrary choice. As we size by 40% (0.4 dex). Here, the the total emission is dominated discussed before, we simply measure the distance to all the spax- by unresolved emission (C< 0.2) and is most strongly affected els that are above the surface brightness threshold. Among those by the beam smearing.

spaxels, the one with the greatest distance defines the ENLR If we look at the power-law slope of the outer surface bright-size. Apparently, the ENLR dominates the emission at large radii ness profile between 1""2.5"" away from the QSO, we find

(9)

0.0 0.2 0.4 0.6 0.8 1.0

d

QSO

(total) [kpc]

0.0

0.5

1.0

1.5

2.0

2.5

d

QSO

(ENLR)

[kpc]

5

10

15

20

R

int

(total) [kpc]

5

10

15

20

R

int

(ENLR)

[kpc]

0

400

800

1200

W

80

(total) [km

/

s]

0

400

800

1200

W

80

(ENLR)

[km

/

s]

0.0 0.1 0.2 0.3 0.4

e

(total)

0.0

0.1

0.2

0.3

0.4

e

(ENLR)

1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

η

(total)

1

0

1

2

3

4

5

η

(ENLR)

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

contrast ratio (

C

)

Fig.7.Comparison of the total and ENLR [O

iii

] line measurements for the centroid distance from the QSO (dQSO), maximum size out to a fixed

surface brightness limit (Rint), the median [O

iii

] line width within the central 1""(W80), the ellipticity of the central [O

iii

] emitting region (e) and

the outer power-law radial surface brightness slope (η). The 1:1 relations are shown as a dashed line in each panel. The colour of each data point corresponds to the contrast ratio Cas indicated by the colour bar which is defined as the fraction of resolved to the total [O

iii

] emission.

wings of the PSF still contribute to the surface brightness be­ yond 1"" making the profile steeper. At the lowest contrast ra­ tio, SDSS J0924+0624 stands out again, because the size of the ENLR is much smaller than 2.5""after subtracting the unresolved emission. In this case, the slope actually does not make sense as it is dominated by noise over most of the range. Therefore, we think that the power-law slope of the surface brightness is an ill-defined quantity if the beam smearing is not taken into account for data at the given spatial resolution.

3.3. Spatially resolved kinematics

In Table 2, we also report the characteristic ENLR [O

iii

] line width W80 as the median of all individual spaxel measurements

within <0"".6 around the QSO position. Three QSOs in the sam­

ple appear to show an [O

iii

] line width of W80 >1000 km s−1

on kpc scales consistent with Liu et al. (2014). We find that in all those cases, the unresolved NLR is at least as bright as the entire ENLR with C< 0.5 and that the [O

iii

] line width in the ENLR reduces significantly to W80 < 800 km s−1 after the deblend­

ing of the NLR. The most extreme difference between NLR and ENLR kinematics is observed for the QSO SDSS J0304+0022 with W80 ∼ 1500 km s−1 for the NLR and almost completely

quiescent kinematics for the ENLR with W80 ∼ 400 km s−1 .

The opposite happens for QSO SDSS J2214+2115 for which we detect significantly broader lines in the ENLR after remov­ ing the NLR contribution. In all the other cases the line widths are either fully consistent with each other or slightly smaller by 100−200 km s−1 .

Depending on the contrast ratio, we see more detailed struc­ ture in the velocity field after the QSO-host deblending. An ex­ treme case is SDSS J0924+0642 (C< 0.2) where we see a sym­ metric velocity gradient across the nucleus with an amplitude of

±230 km s−1. The velocity field in the total light appears flat,

because the velocity of the unresolved NLR dominates over a significant area due to the seeing. The signature for symmet­ ric velocity gradients is also clearly enhanced in the case of SDSS J0304−0707, SDSS J0753+3153 and SDSS J2214+2115. Whether those gradients are due to ordered rotation of a gas disc, or indicate bipolar outflows is unclear at this point.

Another special case is SDSS J0304+0022 for which we de­ tect a huge offset of >500 km s−1 in the radial velocity close to the QSO position after subtracting the unresolved emission. The [O

iii

] line in this QSO has an exceptionally broad blue-shifted component, but only the narrow [O

iii

] component is actually spatially resolved. Thus, the radial velocity measured from the total light is dominated by the unresolved emission up to the ra­ dius where the ENLR emission starts to dominate the [O

iii

] line shape.

3.4. Spatially-resolved line ratios

A key diagnostic for the ionization conditions of the ENLR is the [O

iii

]/Hβ line ratio. Liu et al. (2013a) measured the line ra­ tio across the ENLR for their sample of obscured QSOs as a function of [O

iii

] surface brightness and distance from the QSO. They reported an almost constant ratio [O

iii

]/Hβ ∼ 10 up to a characteristic radius of R ∼ 7 kpc after which the line ratio drops continuously. Here, we present the same analysis for the unobscured QSOs that was not presented in Liu et al. (2014). In Fig. 8, we show the [O

iii

]/Hβ line ratio for all the spaxels with a S/N > 5 before and after subtracting the unresolved emission contribution. The imposed S/N condition leads to a detection limit for the line ratio that varies with the [O

iii

] surface bright­ ness given the fixed depth of the given dataset.

(10)

16.5 16.0 15.5 15.0 14.5 14.0 13.5

log(Σ

[OIII]

/

[erg s

−1

cm

−2

arcsec

−2

])

0.1

0.3

0.5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

log([OIII]

λ

5007

/

H

β

)

3.0

3.5

4.0

4.5

log(

D/

[pc])

SDSSJ0233-0743 (

z

=0

.

455

)

1

00

ENLR only

QSO+ENLR

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

log([OIII]

5007

/

H

β

)

16.5 16.0 15.5 15.0 14.5 14.0 13.5

log(Σ

[OIII]

/

[erg s

−1

cm

−2

arcsec

−2

])

0.1

0.3

0.5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

log([OIII]

λ

5007

/

H

β

)

3.0

3.5

4.0

4.5

log(

D/

[pc])

SDSSJ0304+0022 (

z

=0

.

642

)

1

00

ENLR only

QSO+ENLR

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

log([OIII]

5007

/

H

β

)

16.5 16.0 15.5 15.0 14.5 14.0 13.5

log(Σ

[OIII]

/

[erg s

−1

cm

−2

arcsec

−2

])

0.1

0.3

0.5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

log([OIII]

λ

5007

/

H

β

)

3.0

3.5

4.0

4.5

log(

D/

[pc])

SDSSJ0311-0707 (

z

=0

.

634

)

1

00

ENLR only

QSO+ENLR

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

log([OIII]

5007

/

H

β

)

16.5 16.0 15.5 15.0 14.5 14.0 13.5

log(Σ

[OIII]

/

[erg s

−1

cm

−2

arcsec

−2

])

0.1

0.3

0.5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

log([OIII]

λ

5007

/

H

β

)

3.0

3.5

4.0

4.5

log(

D/

[pc])

SDSSJ0412-0511 (

z

=0

.

549

)

1

00

ENLR only

QSO+ENLR

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

log([OIII]

5007

/

H

β

)

16.5 16.0 15.5 15.0 14.5 14.0 13.5

log(Σ

[OIII]

/

[erg s

−1

cm

−2

arcsec

−2

])

0.1

0.3

0.5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

log([OIII]

λ

5007

/

H

β

)

3.0

3.5

4.0

4.5

log(

D/

[pc])

SDSSJ0753+3153 (

z

=0

.

494

)

1

00

ENLR only

QSO+ENLR

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

log([OIII]

5007

/

H

β

)

(11)

The deblending of the NLR and ENLR does not change the line ratios in most cases and confirms that the ENLR is photoion­ ized by the AGN out to large distances. Exceptions from the flat distributions are SDSS J0233-0743 and SDSS J2214+2115, which show a systematic decrease of [O

iii

]/Hβ at low surface brightness. This decrease can be explained either by ionization of young stars in star forming regions, slow shocks <500 km s−1 in the ISM or if photoionization by the AGN changes from the “ionization” to the “matter-bounded” region. The latter scenario has been favoured by Liu et al. (2013a) as an explanation for the strong decrease in log([O

iii

]/Hβ) after a well-defined break radius.

One important aspect to consider here are the actual detec­ tion limits for the lines. When Σ[O iii] decreases, the Hβ line may

already be below the detection limit depending on the intrinsic line ratio. Since only spaxels are considered for which both lines are detected with >3σ confidence, a bias is introduced towards low [O

iii

]/Hβ line ratios with decreasing Σ[O iii] if low line ratios

are present in the data. It is unclear at this point which role this effect plays in the analysis of the corresponding obscured QSOs sample (Liu et al. 2013a,b).

4. AGN outflow energetics

The estimation of the ionized gas outflow energetics and mass outflow rate is a difficult task and usually depends on assump­ tion on parameters that are not directly constrained by the data. In particular, the lack of spatial resolution usually does not allow us to directly constrain the geometry of the ionized gas outflows. This is even worse for high-redshift AGN where the spatial res­ olution is limited to a few kpc per resolution element. Here, we primarily focus on the comparison of estimates from different models and evaluate how strongly they are affected by contribu­ tions from an unresolved source owing to beam smearing.

4.1. Ionized gas mass and kinetic energy

The amount of ionized gas is set by the amount of ion­ ized hydrogen which can be estimated from the photons emit­ ted by the recombination lines. Adopting “case B” recom­ bination for the low-density limit and a gas temperature of 10 000 K, we expect an intrinsic Balmer line decrement of Hα/Hβ = 2.85 (Osterbrock & Ferland 2006). The ionized gas mass can then be approximated from the Hβ luminosity follow­ ing (Osterbrock & Ferland 2006) as

1.4mp 100 cm−2 LHβ

Mion =

neαeff

LHβ= 107 M0

hνHβ ne 1041 erg s−1 Hβ

(1) where mp is the proton mass, ne is the electron density and h is

the Planck constant. Although Hβ is covered in the observed wavelength range, it suffers from a much lower S/N per reso­ lution element. We therefore use the bright [O

iii

] as a surrogate for Hβ with a line ratio of [O

iii

]/Hβ∼10. Adopting this fixed line-ratio is accurate within ±0.2 dex for all objects as verified by the line ratio distribution (Fig. 8). In fact this provides a lower limit for the ionized gas mass as we do not apply any correction for internal dust extinction.

The greatest uncertainty in our case is the unconstrained electron density ne. We cannot infer it from the data itself be­

cause no density-sensitive lines are in the covered wavelength

range. For the following calculations, we adopt an electron den­ sity of ne ∼ 100 cm−2 as a reference. This value is a typi­

cal value observed in the ENLR around luminous QSOs (e.g. Husemann et al. 2016). However, the density has a large range since it is decreasing with distance (e.g. Bennert et al. 2006b,a) from ne ∼ 1000 cm−2 in the NLR on 100 pc scales (e.g.

Vaona et al. 2012) and ne ∼ 10 cm−2 in the very extended and

diffuse medium on kpc scale (e.g. Liu et al. 2013b).

The total kinetic energy of the ionized nebulae is split into bulk motion vg and the turbulent motion σg of the gas as mea­

sured from the emission lines. With the assumption of constant electron density we can simply integrate the localized kinetic en­ ergy per spatial pixel at position x and y leading to

Ekin =

1

2 Mion(x,y) vg(x,y)

2 +σ

g(x,y)2 . (2)

x,y

The results for the estimated ionized gas mass and the kinetic energy are listed in Table 3. The total ionized gas mass is in the range of 0.6–10 × 108 M

0with a mean of 2 × 108 M0. Al­

though Liu et al. (2014) did not estimate the ionized gas mass for this unobscured QSO sample, they reported a similar ionized gas mass of 6 ×108 M0for their obscured QSO sample (Liu et al.

2013b) following the same assumptions.

Here, we derive the kinetic energy from the kinematics of the total and the ENLR [O

iii

] line profile distribution across the field separately. We note that in the majority of cases the kinematic energies do not change significantly. Only QSOs with a lower contrast ratio C show a clear difference which is caused by a lower ionized gas mass in the ENLR and a lower line width on kpc scales which reduces the turbulent energy term. We obtain a mean kinetic energy of 10 ×1055 erg and 6 ×1055 erg, for the total

and the ENLR energy, respectively. This is about 1 dex lower than reported by Liu et al. (2013b) for the unobscured QSOs, because they assumed a constant outflow velocity of 760 km s−1 across the entire nebulae.

Assuming a timescale τ for the kinetic energy injection one can roughly estimate a kinetic power E˙kin = Ekin/τ. Usually a

timescale of about 107 yr is assumed for the life time of a lu­ minous QSO phase, but Schawinski et al. (2015) recently sug­ gested a much shorter timescale of 105 yr. This yields kinetic powers of 2 × 1041 erg s−1 and 2 × 1043 erg s−1, respectively, for

the two timescales.

As argued in many studies (e.g. Nesvadba et al. 2006; Cicone et al. 2014), these estimates are lower limits to the ac­ tual kinetic energy, because the geometry and projection effect are not taken into account. Therefore, models for the outflow have been used to improve the estimates. Below we describe the results we obtain for a spherical symmetric outflow Liu et al. (2013b) and a conical outflow model (e.g., Cano-Díaz et al. 2012; Cresci et al. 2015).

4.2. Kinetic power and mass outflow rate

4.2.1. Spherical symmetric outflow model

Liu et al. (2013b) adopted a spherical and symmetric outflow ge­ ometry to estimate the energetics and outflow rate for the ion­ ized gas around luminous obscured QSOs. They argued that a spherical geometry is strongly supported by the apparently round ENLR with almost constant broad lines out to kpc distances. They define a shell at distance D through which they estimate the current kinetic power ( E˙kin) and mass outflow rate M˙ . They

(12)

Table 3. Derived outflow energetics for different models.

log Mion log(Ekin/[erg]) log(E˙kin(D)/[erg s−1])a log(M˙out(D)/[M0yr−1])b

Name Pixel-by-pixel Spherical Biconical Spherical Biconical

tot tot ENLR tot ENLR tot ENLR NLR tot ENLR tot ENLR NLR

SDSS J0233-0743 8.3 55.3 55.3 43.9 43.0 42.5 41.8 42.6 2.3 1.7 1.7 1.1 1.6 SDSS J0304+0022 8.1 56.8 55.3 45.4 43.0 43.8 41.7 43.8 2.8 1.6 2.0 0.9 1.9 SDSS J0311-0707 8.1 55.5 55.5 44.9 43.8 43.3 42.3 43.3 2.6 1.9 1.8 1.1 1.8 SDSS J0412-0511 8.9 55.9 56.4 45.8 44.1 44.4 42.7 44.5 3.3 2.2 2.7 1.4 2.7 SDSS J0753+3153 7.8 54.4 54.9 42.7 41.7 41.2 40.4 41.2 1.6 0.9 0.9 0.2 0.8 SDSS J0809+0743 8.5 56.2 55.5 45.2 43.7 43.5 42.2 43.5 2.9 1.9 2.0 1.0 2.1 SDSS J0847+2940 7.9 54.2 54.8 43.2 42.9 41.7 41.4 42.6 1.9 1.7 1.1 0.9 1.1 SDSS J0909+3459 8.3 54.7 54.8 44.2 43.7 42.6 42.2 43.1 2.5 2.2 1.6 1.3 1.6 SDSS J0924+0642 8.1 55.6 55.6 44.7 43.2 43.2 41.9 43.5 2.5 1.4 1.8 0.7 1.9 SDSS J0935+5348 8.4 55.6 56.0 44.5 43.6 42.8 42.0 43.0 2.7 2.1 1.7 1.2 1.8 SDSS J1144+1043 8.6 55.9 56.0 44.6 44.0 42.9 42.5 43.3 2.7 2.4 1.8 1.5 1.8 SDSS J2214+2115 7.9 54.9 55.6 43.6 43.7 42.2 42.3 42.4 2.0 1.8 1.3 1.0 1.3

Notes. (a) Kinetic power dreived from the total light and ENLR distribution for distance D=5 kpc from the QSO for the spherical and conical

model. For the conical model we also estimate the kinetic power for the unresolved NLR adopting a distance of D=1 kpc. (b) Mass outflow rate

from the total light and ENLR distribution for distance D=5 kpc from the QSO for the spherical and conical model. For the conical model we also estimate the kinetic power for the unresolved NLR adopting a distance of D=1 kpc.

[O

iii

]/Hβ line ratio as a function of radius which they associate with the transition from ionization- to matter-bounded clouds, which occurs at D ∼ 7 kpc for their sample of QSOs. Based on the assumptions of matter-bounded and pressure-confined clouds, spherical symmetry and ionization equilibrium, Liu et al. (2013a) derived the following relations

˙

Ekin(D) ΣHβ(D)

=

2.6 × 1040 erg s−1 10−15 erg s−1 cm−2 arcsec−1 3

100 cm−3 v

out kpc

× (3)

ne 100 km s−1 D

where ΣHβ(D) is the Hβ surface brightness, corrected for the sur­

face brightness dimming with redshift, at distance D from the QSO, vout is the outflow velocity and ne is the electron den­

sity. The corresponding mass outflow is then defined as M˙ = 2E˙kin/v2 out which corresponds to

˙ Mout(D)

0.08 M0yr−1 =

ΣHβ(D)

10−15 erg s−1 cm−2 arcsec−1

100 cm−3 v

out kpc

× · (4)

ne 100 km s−1 D

Since the maximum ENLR size drops to Rmax = 6 kpc after sub­

tracting the unresolved emission for SDSS J0924+0642 and no clear break radius in the line ratios is detected for any of the objects, we adopt a fixed radius of D= 5 kpc for which we com­ pute the energetics. In this way we can consistently measure the mean [O

iii

] surface brightness within 4 kpc <R< 6 kpc for all objects and therefore achieve comparable estimates among the sample considering the similar luminosity of all QSOs. Given the low spatial resolution of the data there is no objective criterion to adjust the radius on an object-by-object basis for a comparative study.

In the spherical symmetric outflow model, Liu et al. (2013b) predicted the line shape to vary across the field as a function of the distance from the QSO and measured radial velocity vzand adopted a power-law function for the radial luminosity distribu­ tion in [O

iii

] with slope α=η+ 1,

I(D,vz) ∝ (1 − (vz/vout)2)0.5(α−3)D(1−α). (5)

Such a line shape parametrization implies that W80 ∼ 1.3 ×vout

for a power-law slope η ∼ 3.5 and W80 ∼ 1.5 × vout for

a power-law slope η ∼ 2.6 which are the mean slopes for the total and ENLR radial profiles, respectively. In Table 3 we report the computed kinetic powers and mass outflow rates based on the prescription above and adopting an electron den­ sity of ne = 100 cm−3 . We find a mean kinetic power of

E˙kin(D = 5 kpc) = 1045 erg s−1 and mass outflow rate of −1

M˙out(D = 5 kpc) = 450 M0 yr for the initial values and

E˙kin(D = 5 kpc) = 6 × 1043 erg s−1 and mass outflow rate of

M˙out(D =5 kpc) = 100 M0yr−1 for the ENLR after subtract­

ing the unresolved emission contribution, respectively. Thus, the difference is more than an order of magnitude for the kinetic power and a factor of four in the mass outflow rate, but strongly depends on the contrast ratio for each individual source. In the extreme case, the kinetic power drops by more than 2 dex and 1 dex in the mass outflow rate.

The changes in the energetics we state above are only valid for the kinetic power and mass outflow rate going through a sphere at a distance of D = 5 kpc. The lower rates are ex­ pected because of subtraction of the point-like component which leads to a lower mass and smaller outflow velocity at that ra­ dius. However, the decomposition into unresolved and resolved emission also implies that outflow power and mass outflow rate may change with time/distance from the nucleus in particular on scales smaller than 1 kpc. Therefore, it would be important to es­ timate also the outflow power in the unresolved component. By design this is impossible for this specific spherical outflow model as it requires to compute the emission-line surface brightness at a given radius. We therefore explore this difference between the NLR and ENLR energetics details based on a simple bi-conical outflow model below.

4.2.2. Simple conical outflow model

(13)

AGN based on HST long-slit spectroscopy. It is not clear at this point if this is due to misaligned slits, weak/small outflows or a different geometry. On one hand, a conical outflow geometry is a natural outcome of the unified AGN model that can be eas­ ily resolved and confirmed for many nearby AGN with HST, if present. On the other hand, the opening angle is expected to in­ crease with AGN luminosity so that for luminous QSOs a quasi-spherical outflow cannot be ruled out.

The spatial resolution of these luminous AGN at higher red-shift does not allow us to directly constrain the geometrical pa­ rameters for this model. Cano-Díaz et al. (2012) and Cresci et al. (2015) preferred a conical outflow geometry and adopted a sim­ ple model for their high-z QSOs. The authors assumed a coni­ cal geometry with opening angle Ω, uniformly distributed clouds with the same density and a constant outflow velocity. With the assumption of constant density clouds, the kinetic power and mass outflow rate become independent of the opening angle and the filling factor of the clouds within the cone and one derives the following relations:

3 Mionvout 3

E˙kin(D) = (6)

2 D Mionvout

M˙out(D) = 3 · (7)

D

Assuming case B recombination with an electron temperature of T ∼ 104 K, we can replace M

ion with Eq. (1) which leads to 3

E˙kin(D) = 100 cm−3 LHβ vout kpc

1040 erg s−1 n

e 1041 erg s−1 100 km s−1 D

(8) M˙out(D) 100 cm−3 LHβ vout kpc

= ·

3 M0yr−1 ne 1041 erg s−1 100 km s−1 D

(9) To be consistent with the estimates based on the spherical out­ flow model, we measure the [O

iii

] luminosity within D = 5 kpc converted to Hβ luminosity with a factor of 0.1 and as­ sume again an electron density of ne = 100 cm−3 . The out­

flow velocities are assumed to be maximum velocities in the works of Cano-Díaz et al. (2012) and Cresci et al. (2015) so that we assume W80 to be the representative outflow veloc­

ity. In this case, we obtain kinetic powers and mass outflow rates as reported in Table 3 with a mean kinetic power of

˙ 3 × 1043 −1

Ekin(D = 5 kpc) = erg s and mass outflow rate

˙ −1

of Mout(D = 5 kpc) = 85 M0yr for the initial values and

E˙kin(D = 5 kpc) = 2 × 1042 erg s−1 and mass outflow rate of

M˙out(D = 5 kpc) = 16 M0yr−1 for the ENLR after subtract­

ing the contribution of unresolved emission, respectively. The change owing to the beam smearing of the unresolved NLR is again about 1 dex for the kinetic power and a factor of five in mass outflow rate. Furthermore, the mean kinetic power and out­ flow rate is almost two orders of magnitude lower in the conical compared to the spherical outflow model.

In the same way, we can roughly estimate the kinetic power and mass outflow rate at smaller distances from the unresolved NLR component. Consequently, we adopt the assumptions for the unresolved NLR of a cone with a size of 1 kpc, a slightly higher density of ne ∼ 600 cm−3 and a corresponding outflow ve­

locity based on the [O

iii

] line width in the unresolved QSO spec­ trum. Given that the assumed simple conical model is indepen­ dent of the opening angle by design, we also consider it valid for an approximation for a spherical model given that we cannot

constrain the outflow geometry for the unresolved NLR. Under these assumptions we obtain up to an order of magnitude higher kinetic power and outflow rates in the compact NLR than in the ENLR (see Table 3). While this may indicate much more power­ ful outflows close to the nucleus these values have to be taken skeptically. The assumption of constant-density clouds across the cone is expected to be strongly violated on these small scales and can vary by orders of magnitude up to several 1000 cm−3 . Without spatially resolving the electron density via density-sensitive emission-lines, like done for local Seyfert galaxies with HST (e.g. Crenshaw & Kraemer 2000; Rice et al. 2006; Storchi-Bergmann et al. 2010; Fischer et al. 2013), no firm con­ clusions can be made. However, it is certainly possible that a high mass outflow rate is still confined to a region less than 1 kpc and has not travelled throughout the host galaxy yet.

5. Discussion

5.1. Morphology and incidence of kpc scale outflows

A major result of the work by Liu et al. (2013b, 2014) is that the ENLR appears round with a constant line width of W80 ∼

Figure

Updating...

Related subjects :